TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 12, No. 10, Octobe
r 20
14, pp. 7318
~ 732
9
DOI: 10.115
9
1
/telkomni
ka.
v
12i8.649
3
7318
Re
cei
v
ed Ma
y 16, 201
4; Revi
sed
Jul
y
1
5
, 2014; Acce
pted Augu
st 10, 2014
Intelligent Linear Collaborative Beamforming for Multi-
objective Radiation Beampattern in Wireless S
e
nsor
Networks
N.N.N.
A. Malik*, M. Esa,
S.
K.S. Yusof, N.M.A. Latif
f
UT
M MIMOS
CoE T
e
lecom
m
unci
a
tion T
e
chno
log
y
, F
a
cu
lty of Electric
al
Engi
neer
in
g,
Univers
i
ti T
e
knolo
g
i Mal
a
ysia
(UT
M
),
81310
UT
MJB, Johor Bahru, Mal
a
ysi
a
.
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: noord
i
ni
@fke
.utm.my
A
b
st
r
a
ct
Coll
ab
orative
b
e
a
m
for
m
i
ng (
C
B) in
w
i
reless
s
ensor
netw
o
r
ks (W
SNs) pro
m
is
es i
m
prov
e
m
e
n
t of
communic
a
tio
n
perfor
m
ance
and
en
ergy
effi
ciency. T
h
e ra
ndo
m distri
buti
on s
ensor
n
o
d
e
s l
o
catio
n
w
i
thi
n
W
S
Ns can
int
r
oduc
e ra
nd
o
m
be
ampatter
n
mostly
in t
h
e sid
e
l
obe
re
gio
n
. In a
d
d
i
ti
on, h
i
gh
er e
n
e
rgy
consu
m
ption c
an occur
as th
e rand
o
m
n
e
ss per
mits the
g
e
nerati
on of h
i
g
h
peaks
in ra
di
ation b
e
a
m
patter
n
perfor
m
a
n
ce. T
herefore, sel
e
cting a
suita
b
l
e
spatia
l sens
or nod
e distrib
u
tion is a cha
l
l
eng
e espec
ia
ll
y for
W
S
Ns. Coll
ab
orative s
ens
or
nod
es i
n
ra
ndo
m
dep
loy
m
ent
w
h
ich p
e
rforms as li
ne
ar a
n
tenn
a arr
a
y (LA
A
)
can infl
ue
nce
the radi
atio
n
bea
mpattern.
How
e
ver,
it l
eads to th
e d
egra
datio
n of
LAA an
d W
S
Ns
perfor
m
a
n
ces.
Henc
e, an
opti
m
u
m
al
gorit
hm for i
m
p
l
e
m
e
n
ti
ng CB
meth
od
sho
u
ld
be
des
ign
ed th
at take
s
into cons
id
erat
ion n
o
t only th
e bea
mpattern
performanc
e, but also th
e g
e
o
m
etric
a
l loc
a
tion of sel
e
ct
ed
active n
odes
w
h
ich coo
pera
t
ively for
m
an
array ant
enn
a. In this artic
l
e, a n
e
w
alg
o
rith
m kn
ow
n
a
s
intell
ig
ent l
i
ne
a
r
sensor
no
de
array (ILSA)
i
s
pres
e
n
ted. It is dev
el
ope
d
throug
h the
a
pplic
atio
n of t
h
e
prop
osed
hy
br
id l
east s
quar
e i
m
pr
ove
d
p
a
rticle
sw
ar
m opti
m
i
z
a
t
io
n
(HLPSO) al
go
rithm. T
h
e n
e
w
ly
prop
osed IL
SA
is constructe
d
by means
of colla
bor
ative n
o
des sel
e
ctio
n. T
he si
z
e
of si
d
e
lob
e
l
e
vel (S
LL)
can vary s
i
g
n
i
ficantly w
i
th
desir
ed
mu
lti-
obj
ective
s. Si
mu
lati
on res
u
l
t
s obtain
ed s
how
ed si
gn
ific
ant
improve
d
perf
o
rmanc
e of the radi
ati
on b
e
a
mpattern. T
h
us, this motiv
a
tes for explo
i
ting the n
e
w
l
y-
deve
l
op
ed o
p
ti
mu
m
meth
od i
n
nod
e ge
o
m
et
rical loc
a
tio
n
strategi
es of W
S
Ns.
Ke
y
w
ords
: col
l
ab
orative b
e
a
m
for
m
i
ng (CB),
line
a
r ante
nna
array (LAA), wireless s
ensor
netw
o
rk (W
SN)
Co
p
y
rig
h
t
© 2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g an
d
Scien
ce. All righ
ts reser
ved
.
1. Introduc
tion
As the de
ma
nd for
wirele
ss sen
s
or
net
works
(WSNs) co
nsta
ntly grows, the
req
u
irem
ent
for hig
her transmi
ssion
range, b
e
tter cove
ra
g
e
and im
prove
d
communi
cation reliabili
ty
increa
se
s. Typical si
gnal
transmi
ssio
n issue re
qu
ires
sufficie
n
t gain by using a high g
a
in
antenn
a ele
m
ent or
con
s
tru
c
ting a
n
array of el
e
m
ents that o
perate
co
herently. However, in
WSN e
n
viro
n
m
ent, the ra
ndomn
e
ss g
eometri
cal lo
cation
of sen
s
or
nod
es i
n
transmitting
an
emitted sig
n
a
l
pre
s
ent
s a
uniqu
e situ
ation. T
he
den
se d
eployme
nt of many n
ode
s would
not
allow
co
stly and
com
p
lex high g
a
in
antenn
a to
be u
s
ed. T
h
erefo
r
e, u
s
in
g the p
o
we
r of
beamfo
rming
in the sen
s
o
r
node
arrays is an excell
e
n
t alternative. Individual se
nso
r
nod
e ha
s
an om
nidirect
ional a
n
tenn
a
,
which radiat
es
p
o
wer
unif
o
rmly in
all di
rectio
ns, i.e.
360
. Ho
wev
e
r,
most of th
e transmitted
po
wer is waste
d
while
only
a
fraction
of the
tran
smitted
p
o
we
r i
s
u
s
efu
l
,
whi
c
h propa
g
a
tes to the d
e
sired di
re
ction. Therefore
,
if multip
le sensor no
de
s colla
borate a
nd
coo
r
din
a
te th
eir tra
n
smi
s
si
ons
by se
ndi
ng the
sa
me
sign
al,
the
propag
ating sig
nals will
inte
rfere
constructively. Thus, the t
o
tal ra
diated
power
will increase and f
o
cus to the desi
red directi
on.
Furthe
rmo
r
e,
collab
o
rative
beamformin
g (CB
)
in WSNs
can effe
ctively impro
v
e reliability and
coverage by
nulling patt
e
rn
s towa
rd
s interfere
r
s,
and the tran
smissio
n
ra
n
ge will al
so
be
increa
sed.
Re
sea
r
ch o
n
se
nsor net
works u
s
in
g
CB [1
-4] hav
e explo
r
ed
the effe
ctiven
ess a
nd
compatibility of the transm
issi
on ar
ray on the random
ly distribut
ed
sensor nodes. The effect
of
rand
om moti
on by indepe
ndent mobil
e
sen
s
or n
ode
s ha
s been d
i
scusse
d in [5]. The propo
sed
relation
shi
p
model
ha
s m
anag
ed to
re
duce n
e
two
r
k overh
ead
an
d po
we
r u
s
a
ge in
a
wirel
e
ss
netwo
rk
stru
cture. Ref. [6] refined a
sp
e
c
ific an
al
ysi
s
on focu
sin
g
e
nergy in a d
e
s
ire
d
directio
n
usin
g a
comb
ination of tim
e
differen
c
e
of arrival
(T
DOA) an
d ad
a
p
tive beamfo
rming te
chniq
ue.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Intelligent Lin
ear Collab
o
ra
tive Beam
form
ing fo
r Multi-obje
c
tive Ra
diation… (N.
N
.N.A. Malik)
7319
The en
ergy
saving
s meth
odolo
g
y has
also
been i
n
vestigated in
[7] to provid
e better d
e
fined
beam
width
a
nd
com
pen
sate the
effect
of g
r
ating
lo
bes.
Two diff
erent
metho
d
s
a
r
e
p
r
opo
sed
with lea
s
t me
an squa
re
(L
MS) implem
e
n
tation. Th
e
s
e wo
rks m
a
in
ly investigate
the pe
rform
a
nce
of beamfo
rmi
ng u
s
ing
the
theory
of ra
ndom
arra
ys.
T
o date, little
attention h
a
s
b
een
paid
to
linear
arrays for CB in li
near
arrays.
Ho
weve
r, alt
e
rnative a
pproache
s have
been lin
ke
d
to
rand
om a
r
ray node im
plem
entation. Works
on the li
ne
ar a
rray
pro
p
o
se
d by [8] h
a
s b
een fu
rth
e
r
develop
ed
an
d repo
rted [9
-12]. In [13], a
n
iterativ
e ap
proa
ch
that
can b
e
e
m
ploy
ed in
sce
n
a
r
i
o
s
whe
r
e th
e in
dividual n
o
d
e
s d
o
n
o
t h
a
ve kn
owl
e
d
ge of thei
r l
o
catio
n
ha
s
been
propo
sed.
Papalexidi
s
[
8
] and
Mali
k [12] utilize
a
lea
s
t sq
ua
re
line fitting te
chni
que
(LFA
) to
sele
ct th
e
optimal nod
e
s
to parti
cipat
e in the distri
buted array.
A challen
ge
with implem
e
n
ting beamfo
rming in
WS
Ns i
s
that the sen
s
o
r
no
des a
r
e
locate
d in
ra
n
dom
distri
buti
on. In thi
s
wo
rk, th
e o
p
timization
of
sen
s
or n
ode
loca
tion p
r
oble
m
i
s
investigate
d
whe
r
e a l
a
rg
e numb
e
r of
node
s a
r
e
available to t
a
ke
part in t
h
is b
eamfo
rming
action. Obvio
u
sly, the arra
y gain perfo
rmance imp
r
ove
s
w
i
th
inc
r
ea
s
i
ng
num
ber of elements
(i.e.
sen
s
o
r
n
ode
s).
Ho
wever,
the no
de
s in WS
Ns
have limitati
ons in e
nergy con
s
u
m
pt
ion,
comp
utation
power
and
communi
catio
n
ra
nge
s,
wh
ich
mea
n
s
that it is in
a
ppro
p
ri
ate fo
r all
available n
o
d
e
s to take p
a
r
t in this CB
action.
Con
s
eque
ntly, selecting o
n
ly suitable no
de
s fro
m
the availabl
e
active cl
ust
e
r (
AC
) to
perfo
rm be
a
m
forming i
s
of re
sea
r
ch
con
c
e
r
n. In
the
prelimi
nary v
e
rsi
on of thi
s
work [11], it wa
s s
hown that the linea
r sen
s
o
r
no
de
s array (LSNA) is
able to achie
v
e a comp
arable ad
aptive beamp
a
tte
rn with na
rro
w main lo
be
and acce
pta
b
le
sidel
obe
s lev
e
l (SLL
). Thi
s
pap
er exte
nds e
a
rli
e
r
work
rep
o
rte
d
in [14, 15]. Novel con
c
e
p
t is
offered in reg
a
rd
s to intelligently optimizing the
sel
e
ct
ed se
nsor no
des to pa
rtici
pate and form an
array of sen
s
or no
de
s by employing int
e
lligent line
a
r sen
s
o
r
nod
e
array (ILSA
)
bas
ed
on
t
h
e
swarm
intelligence al
gorithm
[16].
T
h
is new
algori
t
hm
of ILSA wi
th
combinati
o
n
of novel
hybri
d
le
ast
sq
uare
improve
d
-PSO
(HLPSO) h
a
s
bee
n pr
op
osed
t
o
s
u
it
th
e WSNs re
quiremen
t
. The
swarm
algorit
h
m has
been
chosen as
the problem
sol
v
er because
of
its flexibilit
y, versatility and
ability to optimize in compl
e
x multimodal search
spaces.
The m
a
in id
ea of the
p
r
opo
sed
meth
od i
s
t
he
de
sire
d m
u
lti-o
b
jective
s
of
radiation
beamp
a
ttern
with minimu
m SLL, desi
r
ed main be
a
m
angle an
d
size of first null beam
wi
dth
(FNB
W). T
h
e
pro
p
o
s
ed i
n
telligent meth
od of ILSA fo
r dete
r
minin
g
optimum lo
cation of sen
s
or
node i
s
prov
ed su
peri
o
r to alternate t
e
ch
niqu
e in term
s of the norm
a
lized p
o
we
r gain. T
h
e
remai
nde
r of this articl
e is
orga
nized a
s
follows
. Sect
ion 2 provid
e
s
an overvie
w
of the syst
em
modelling and array
factor. Section
3
explai
ns the
HLPSO-based IL
SA method. Section
4
pre
s
ent
s an a
nalysi
s
of the prop
osed met
hod.
The la
st se
ction sum
m
ari
z
e
s
the p
aper.
2. The Sy
ste
m
Model and Arra
y
Factor
WSNs
con
s
i
s
t of a larg
e
numbe
r of
sensor n
ode
s;
whi
c
h a
r
e
wirele
ssly
con
necte
d in
rand
om po
siti
on. The no
de
s are self
-org
anized an
d
a
r
e in conta
c
t with a control
ling station a
s
descri
bed i
n
[3]. Each of this sen
s
or
n
ode’
s lo
catio
n
is dete
r
min
ed usi
ng lo
cation discove
r
y
techni
que
s [7
] and
is repo
rted ba
ck to
the
cont
ro
lle
r. The
controll
er
ha
s d
e
tail
ed
kno
w
le
dg
e of
each of the
sensor
node’
s
locatio
n
. The
s
e n
ode
s the
n
form
a diffe
rent
clu
s
ter
with reg
a
rd
s to
the
sele
cted
man
ager no
de
(
MN
).
Th
e
cont
rolle
r i
s
al
so
cap
able
of
se
lecting
the
ap
prop
riate
MN
,
thus a
c
tive cl
uster
(
AC
) a
s
per u
s
e
r
req
u
irem
ent. Each of the
sen
s
ing
node
s,
S
z
is able to se
nse
the environm
ent and
colle
ct
its o
w
n
dat
a. The
sel
e
ct
ed
MN
g
a
the
r
s th
e d
a
ta from the
se
nsi
ng
node
s an
d th
en multicast
a final data p
a
cket to a
ll the sele
cted
co
llaborativ
e se
nso
r
no
de
s, i.e.
active CB n
o
des. Th
e dat
a from the
s
e
sen
s
in
g nod
e
s
are ag
gre
g
a
ted at the
MN
and
only the
need
ed information will b
e
multica
s
ted
.
The active CB node
s wi
ll collabo
rativ
e
ly transmit the
same
data i
n
a syn
c
h
r
o
nou
s man
ner. These
acti
ve CB nod
e
s
; whi
c
h p
e
rf
orm a
s
an
a
rray
antenn
a hav
e the p
o
ssibil
ity of forming
a na
rrow
hi
ghly directive
beam
to the
intende
d target
point, wh
ere
the re
ceive
r
s
(ba
s
e
sta
t
ions) may
be po
sition
e
d
in o
r
de
r t
o
coll
ect all
the
transmitted d
a
ta sent by collabo
rative n
ode
s.
The
colla
bora
t
ive array ant
enna
ra
diate
s
po
wer in
all
dire
ction
s
, th
erefo
r
e it i
s
a
s
sumed
that all
se
nsor
nod
es a
r
e lo
cated
on
a
3-di
men
s
i
onal
x-
y-
z
pl
ane. T
he
ge
ometri
cal
mo
del
config
uratio
n of
the ran
d
o
m
ly
depl
oym
ent sen
s
or n
o
des an
d the
target
point
is
illustrate
d in
Fig.
1. As
sho
w
n,
there
are
thre
e differe
nt set
s
of
sen
s
o
r
n
ode
s (i.e. a
c
ti
ve/transmit m
ode, idle
mod
e
and
sle
ep
mo
de). A
c
tive m
ode i
ndi
cate
s a
node
in
a
c
tive state
and
tran
smitting
colle
cted
dat
a
.
Idle mode ind
i
cate
s a nod
e
in active state but waiting
to transmit da
ta. The sleep
mode indi
cat
e
s
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 10, Octobe
r 2014: 731
8
– 7329
7320
a node in
sle
ep state
with very minimal
energy usa
g
e
.
Both active and idle n
ode
s are within t
h
e
AC
while the
slee
p mode n
ode
s are o
u
tside the
AC
.
The
pro
p
o
s
e
d
sy
stem i
s
forme
d
by
a
netwo
rk of
Z
station
a
ry
sensor
no
des
rand
omly
placed at po
sition
S
z
= (
s
1
, s
2
,..,s
Z
). Each
sen
s
o
r
nod
e
is den
oted in
Carte
s
ia
n co
ordin
a
te (
x
k
,y
k
)
with
k
denot
es the
numb
e
r of no
de
s
while the l
o
cati
on of targ
et point is gi
ven in sp
he
ri
cal
c
o
or
d
i
na
te
s
,
P
(
p,
0
,
,
0
)
,
where
p,
0
,
an
d
0
,
are the
distan
ce
bet
wee
n
the ta
rget point
and
the
referen
c
e po
int, desire
d
elevation an
d azimuth a
ngle, re
spe
c
t
i
vely. The a
n
gle
[-
,
]
rep
r
e
s
ent
s th
e a
z
imuth
direction
an
d
[0,
] repre
s
ent
s the
ele
v
ation directi
o
n. The
follo
wing
assumptio
n
s
are ad
opted i
n
con
s
id
erin
g
this model o
p
eratio
n:
(a)
Senso
r
n
ode
s a
r
e a
s
sum
e
d to lie in
the
x-
y
plane
in
rand
om d
epl
oyment in
sid
e
the
regio
n
of interest of
Λ
m
2
.
(b) M
u
tual-couplin
g effects am
ong t
he anten
na
s of different sen
s
o
r
no
des a
r
e
negligibl
e
. No
signal reflecti
on or scatte
ri
ng,
thus multi
path fading o
r
shad
owi
ng.
(c) T
he lo
cati
on of ta
rget
points,
P
is
at the far-field region from the
AC
with
de
sired
angle of
0
and
0
.
(d) All nod
es
are statio
na
ry and ene
rgy constraine
d.
(e) All nod
es
are capa
ble o
f
operating a
s
a
MN
either i
n
active, idle and sl
eep mo
des.
Figure 1.
Definition of Notation
Con
s
id
er
a 3
-
dime
nsi
onal
cha
r
a
c
teri
stic
of
N-
e
l
e
m
ent L
AA p
l
ac
ed a
t
th
e
x-y
-
z
plan
e
.
A
ssu
me
z
=
0, therefore the plane i
s
visuali
z
e
d
to ru
n parall
e
l to the earth’
s
su
rface. Th
e array
fac
t
or,
AF
of the LAA is ref
e
rred by [17]:
,
∑
(1)
Both the current sign
al ph
ase,
and th
e synchro
n
izi
ng pha
se
wei
ghts,
can b
e
determi
ned
by:
(
2
)
(
3
)
W
h
er
e
,
θ
,
,
x
n
an
d
y
n
are
th
e wa
ve
nu
mb
er
= 2
/
λ
with
λ
i
s
the
wavel
e
ngth,
elevation
ang
l
e,
azimuth angl
e,
x
-coo
rdin
ate and
y
-coo
rdinate (
x
n
,y
n
) of
the
n
th element, res
p
ec
t
i
vely.
0
and
0
are the maxi
mum radi
atio
n angle
s
.
The po
we
r ga
in
G
is then gi
ven by [17]:
,
20log
|
,
|
(
4
)
And the norm
a
lize
d
po
wer
gain
G
norm
in
deci
bel [17]:
,
1
0
l
o
g
|
,
|
|
,
|
(5)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Intelligent Lin
ear Collab
o
ra
tive Beam
form
ing fo
r Multi-obje
c
tive Ra
diation… (N.
N
.N.A. Malik)
7321
3. Proposed
Metho
dolog
y
Description
3.1. Propose
d
H
y
brid Least Squar
e
Improv
ed-PSO Algorithm (HLPSO)
PSO
is
empl
oye
d
to
d
e
t
e
rmine t
he o
p
t
imum dist
an
ce loc
a
ti
on
o
f
th
e
no
de
s;
which
perf
o
rms t
h
e
best wi
thi
n
the
obj
ectiv
e
scop
e
s.
So
me imp
r
ove
m
ents
have
been
ado
pte
d
in
origin
al PSO [16] in orde
r to overcome t
he we
ak
ne
sses an
d to ada
pt the algorith
m
insid
e
WS
Ns
environ
ment.
The n
o
vel HL
PSO is p
r
op
o
s
ed
by ad
opti
ng two
novel
mech
ani
sm
s, i.e. con
s
trai
nt
boun
dari
e
s v
a
riabl
es
and
particl
e’s p
o
s
ition an
d ve
locity reinitial
i
zation. Mo
re
over, the lea
s
t
squ
a
re
app
ro
ximation algo
rithm (LS
)
is
integrate
d
int
o
it to improv
e the effectiv
ene
ss
and th
e
cap
abilities of
PSO in ILSA application.
The
m
odi
ficat
i
ons
an
d
impr
ovements
th
at
hav
e
b
een
do
ne
in
HLPSO
are
discuss
ed
as follows:
3.
1.
1
.
G
l
ob
al
Constr
aint Bound
aries Variables
Two sets
of global con
s
traints
bo
un
d
a
r
ies va
riabl
es for lowe
r bo
unda
ry,
L
an
d uppe
r
boun
dary,
U
for different p
o
sition
pa
rticl
e
s,
d
s
1
a
nd
d
sn
(
n=
2,3,
…N
)
is ado
pted and rep
r
e
s
en
ted
as:
(
6
)
(7)
These two b
ound
arie
s
are ap
plied to
rest
rict
d
s
1
a
nd
d
sn
t
o
st
a
y
inside
t
he
solut
i
o
n
spa
c
e. Ad
ditionally, maxi
mum u
ppe
r li
mit and mi
ni
mum lo
we
r li
mit are
also
assimilated
i
n
sid
e
this propo
se
d
HLPSO, i.e.
U
ma
x
and
L
mi
n
, resp
ectively
. These t
w
o l
i
mits are det
ermin
ed befo
r
e
the comp
utation of the objective functio
n
,
of
in order to enhance the diversity of the particle’
s
sea
r
ching a
b
i
lities to be more glo
bal an
d freedo
m. Thus, it is expressed a
s
:
,
,
,
(8)
And
,
,
,
(
9
)
3.1.2. Particle’s Position a
nd Velocit
y
Reinitializ
ation
The ra
ndo
m numbe
rs of p
a
rticle
po
sitio
n
, d
sn
can be
a factor of th
e parti
cle’
s tenden
cy
to leave the initially define
d
sea
r
ch spa
c
e.
The
r
efo
r
e
,
a modificati
on ba
sed o
n
the abso
r
bi
ng
wall conditio
n
s by [18] is implemente
d
in this al
go
rithm. In ord
e
r to co
ntrol
the moveme
nt of
particl
e from
flying outside
the borde
r
of the search
space, the velocity,
v
sn
is zeroe
d
wh
ene
v
e
r
the pa
rticle,
d
sn
goe
s ove
r
the bou
nda
ry
U
N
and
L
N
. However, the partic
le,
d
sn
are th
en p
u
ll
ed
back in
sid
e
th
e se
arch
spa
c
e by
rein
itiali
zing it as
random numbers,
r
gen
erate
d
from the valu
es
of [
L
mi
n
,
U
ma
x
]. The
obj
ectiv
e
of thi
s
rei
n
itialization
of
d
sn
is to
p
r
e
v
ent the
part
i
cle f
r
om
bei
ng
stucke
d in lo
cal optima
sce
nario
wh
ere the pa
rticle
is
trappe
d an
d i
nhibited to
se
arch for a b
e
tter
solutio
n
. By introdu
cin
g
th
e reinitili
zatio
n
, a more fle
x
ible and
co
mpre
hen
sive
sea
r
ching
ca
n be
done by the p
a
rticle
with no
ted limitat
ions, as expre
s
se
d by equation
s
belo
w
:
0
→
,
,
,
0
→
,
,
(
1
0
)
By using Equ
a
tion (1
0), th
e parti
cle mo
vement mayb
e trigge
red
a
gain so that it has the
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 10, Octobe
r 2014: 731
8
– 7329
7322
highe
r pro
b
a
b
ility to search for the optimum glob
al best. In additio
n
, the particle
position i
s
also
force
d
to sta
y
inside the
uppe
r bo
und
ary,
U
an
d lowe
r bo
und
a
r
y,
L
as d
e
n
o
ted by follo
wing
equatio
ns:
,
,
,
(11)
And,
,
,
,
(
1
2
)
The integ
r
ati
on of the LS
approximatio
n algo
ri
thm in
this HLPSO i
s
re
quired
so
that the
desi
r
ed
ra
di
ation be
amp
a
ttern p
e
rfo
r
mance
c
an
be cl
osely a
pproxim
ated
to the de
si
red
beamp
a
ttern
results. Due to the ran
dom
spatial p
o
siti
oning of n
o
d
e
s, LS algo
rit
h
m provid
es t
he
ability to alter and
create
a radi
ation b
eampatte
rn b
y
introdu
cing
weight
s on
each nod
e. T
he
determi
nation
of the
weigh
t
s allo
ws eli
m
ination
of
the
effect
of ra
n
dom
node
s p
o
sition
erro
rs in
WSNs. The e
ffect of weigh
t
s can b
e
rem
o
ved throu
gh
equali
z
ation.
3.2. Optimization Me
tho
dolog
y
Setup of ILSA
The num
ber
of particip
a
tin
g
CB node
s,
N
is also ini
t
ialized
whi
c
h
are re
pre
s
e
n
ted as
Q
n
= (
q
1
,q
2
,..,q
N
). These CB
node
s a
r
e in
active mod
e
s. The comm
u
n
icatio
n ra
diu
s
,
C
of
MN
ha
s
been ide
n
tified at this stag
e. The di
stan
ce between t
w
o nea
rby CB nodes i
s
/2, which d
e
p
end
s
on the o
perating freq
uen
cy
,
f
. The initial para
m
eters f
o
r
WSNs env
ironm
ent are
sho
w
n i
n
Tab
l
e
1.
The propo
se
d optimizatio
n setup
sche
me of ILSA is
desc
r
ibed as follows
:
Step 1.
Const
r
uct the virtua
l line
that passe
s throu
g
h
the
MN
.
Step 2.
Establish the HLPSO algorith
m
.
Step 2a
. Initialize HLPSO param
e
ter as in Table 2.
Step 2b.
Ra
n
domly gen
erate initial location
D
a
nd v
e
locity
V
for
each pa
rticle
in an
N
-
dimen
s
ion
a
l probl
em with
S
particle
s
:
,
,
,……
∈
0
,
2
(13)
,
,
,…
…
∈
0,
0
.2
(14
)
Step 2
c
. Cal
c
ulate
obj
ecti
ve functio
n
i
.
e.
of
or
m
u
lobj
of
ea
ch
d
sn
, i.e.
of
(
d
sn
) o
r
m
u
lobj
(
d
sn
).
(
1
5
)
∈
(
1
6
)
∙
∙
∙
(
1
7
)
Whe
r
e:
of
SLL
is the objective functio
n
of SLL minimization te
rm
as define
d
in
:
∑
|
|
∑
|
|
(
1
8
)
Wh
er
e
SLL
1
and
SLL
2
a
r
e
th
e
ang
le
s
whe
r
e
the
SLL
is min
i
mized
in
the
lo
wer band
(from
SL
L
1=1
to
SLL
1=
Ma
x
S
L
) and
in
the
uppe
r
b
and
(f
ro
m
SLL
2=
Mi
n
S
L
to
SLL
2
=
181
), respectivel
y
.
AF
is
the ar
ray
fa
cto
r
as in equa
t
i
on
(1
).
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Intelligent Lin
ear Collab
o
ra
tive Beam
form
ing fo
r Multi-obje
c
tive Ra
diation… (N.
N
.N.A. Malik)
7323
of
nu
is the objective functio
n
of null placement term a
s
define
d
in:
∑
|
|
(19)
Where
an
d
nu
a
r
e
t
h
e
n
u
m
b
e
r
o
f
n
u
l
l
s
a
n
d
t
h
e
lo
cat
i
on
ang
le
s
of
nu
ll
p
l
acement
s,
respective
ly.
of
bw
is the objective functio
n
FNBW term as
defined in:
∑|
|
(20)
Wh
er
e
bw
is the
ang
le
of
de
sire
d
FNBW,
i.e.
FNBW=
bw
2
-
bw
1
wh
ich
is the
range
of
ang
les of t
h
e
ma
jo
r lobe.
w
i
, i
=1,2,3
is t
he u
s
e
r-d
efin
ed con
s
tants
tha
t
co
ntrol t
he c
ontri
bu
tion fr
om e
a
ch
term of
sub-
obj
ectiv
e
to t
he
over
a
ll objec
tive f
u
nctio
n
.
It is u
s
ually
assum
ed t
h
a
t
∑
1
and
t
he
value of
w
i
is
determi
ned b
a
se
d on the u
s
er-d
esi
r
ed o
b
jective
s
.
Step 2d
. Determin
e the previous b
e
st locatio
n
,
pbe
st
,
P
=[
p
s
]=
[
p
1
,
,
p
2
,
,
p
3
,….
p
S
].
S
e
t
of
(
p
s
)
value equ
al to
of(d
sn
)
or
m
u
lobj(d
sn
)
.
Step 2e
. Det
e
rmin
e the gl
obal be
st po
sition,
G
=[
g
n
]=
[
g
1
,g
2
,g
3
,..,g
N
]. Set
g
n
=m
in(p
s
)
or
g
n
=optim
um
(p
s
).
Step 2f.
Upda
te
V:
1
1
1
(21
)
W
h
er
e
c
1
and
c
2
are a
c
cel
e
ration
con
s
t
ants,
r
1
a
nd
r
2
are
unifo
rml
y
distri
buted
numbe
rs in
[0
,1].
+1 and
ref
e
r to the time index of the
curre
n
t and p
r
eviou
s
iterati
ons.
is the
inertial weigh
t
factor. Then,
limit
V
using equatio
n (10
)
.
Step 2g.
U
p
date
D
:
1
1
(22)
And limit
D
of the particle
s
by using Equ
a
tion (11
)
an
d (12
)
.
Step 2h
. Update
pbe
st
as
follows
:
If
of(d
sn
)
or
m
u
lobj(d
sn
)
is
better than
of(ps)
or
m
u
lobj(p
s
),
the
n
upd
ate
p
s
and store
d
sn
(
p
s
).
Step 2i.
Upda
te
gbest
as
follows
:
If
of(p
s
)
or
m
u
lobj(p
s
)
is bet
ter than
of(g
n
)
or
m
u
lobj(g
n
)
, then update
g
n
and sto
r
e
d
sn
(
g
n
).
Step 2j
. If
the maximum iteration nu
mbe
r
is met, stop
algorith
m
, else go to
step 2
c
.
Step 3.
By using the
optimi
z
ed
dista
n
ce,
d
sn
obtain
ed
f
r
om previo
us pro
c
e
ss, co
n
s
tru
c
t
a
LAA on th
e
virtual line
,
whi
c
h
is pe
rpendi
cula
r to
the virtu
a
l li
ne
. The
c
o
ns
truc
ted LA
A is
assume
d hav
ing
N
-nod
es with
sp
aci
ng distan
ce
of
d
sn
. The sen
s
o
r
nod
e location of
x
- a
nd
y
-
c
o
or
d
i
na
te
,
E
n
(
x
E
n
, y
E
n
) is
referred to the values
of
d
sn
:
(23)
Whe
r
e
n
=1,
2
,
…
.
N
nodes el
ement. The construction of this
optimum
LAA is illustrated in Fig. 2.
Step 3a
.
Det
e
rmin
e gai
n a
nd no
rmali
z
e
d
gain,
G
E
and
G
norm
E
by using
equ
ation
s
(4)
and
(5), respec
tiv
e
ly.
Step 4
:
The optimu
m
LAA,
E
n
can
not be pe
rformed practi
cal
l
y as the sen
s
or
node
s a
r
e
rand
omly dist
ributed
(unifo
rm ran
dom
di
stributio
n) in
side WSNs fiel
d.
Step 4a
. Se
a
r
ch
the
mini
mum Eu
clide
an di
stan
ce,
d
mi
n
n
between
E
n
(
x
E
n
, y
E
n
) an
d the
nearest no
de
insid
e
AC, O
zs
(x
O
zs
,y
O
zs
).
min
(24)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 10, Octobe
r 2014: 731
8
– 7329
7324
With
zs
=1,2,….
ZS
node
s insid
e
AC
.
Step 4b
. Selec
t
the
O
zs
which ha
s
d
mi
n
n
with co
ordi
na
te (
x
O
zs
,y
O
zs
).
Step 4
c
.
Ac
tivate
O
zs
and assign
it a
s
a
n
optim
um IL
SA. ILSA de
noted
as
S
n
(
x
S
n
, y
S
n
),
S
n
O
zs
. The mappin
g
process is illu
strat
ed in Figu
re 3
.
Figure 2. System Model of the Propo
se
d
ILSA
Figure 3. Locations of
E
n
, O
zs
and
S
n
inside
AC
Step 4d:
This set of optimized
nodes i.e. ILSA will act
collaboratively as an
N
-element
distrib
u
ted L
AA.
Determi
n
ed
g
a
in and
norm
a
lized g
a
in,
G
S
and
G
norm
S
for this
newly
optim
a
l
ILSA by usin
g equ
ation
s
(4)
and
(5
). All these IL
S
A
denote
s
in
active mod
e
,
i.e. active CB
node
s.
Step 5
. Start LS approxim
ation method.
Step 5a.
Calculate the de
si
red weight co
efficient,
w
E
n
e
x
p
(
2
5
)
Step 5b.
Calculate the de
si
red ste
e
ri
ng coefficient,
d
E
nm
e
x
p
(26)
Whe
r
e
is el
evation angl
e
,
{-180°,1
80°} a
nd (
m
=1,2,..361)
Step 5c.
Cal
c
ulate the d
e
s
ire
d
array re
spo
n
se,
F
E
nm
:
(27)
Step 5d.
Calculate the actu
al steeri
ng co
efficient,
d
S
nm
e
x
p
(
2
8
)
Step 5e.
Fro
m
F
E
nm
and
d
S
nm
,
calculate actual weig
ht
coeffici
ent,
w
S
n
(29)
Whe
r
e
F
ET
nm
is the tran
spo
s
e of
F
E
nm
and
d
S+
nm
is the pse
udo inve
rse of the tran
spo
s
e
d
S
nm
.
Step 5f.
Fro
m
w
S
n
, calcul
ate the final
array factor
o
f
S
n
(
x
n
S
,
y
n
S
) active ILSA CB nod
es
as pe
r written
equation:
,
∑
∗
ex
p
(30)
Whe
r
e w
n
*S
is the complex
conj
ugate of
w
n
S
Step 6.
Dete
rmine
gain,
G
ILSA
and no
rmali
z
ed
gai
n,
G
norm
ILSA
of final ILSA as p
e
r
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Intelligent Lin
ear Collab
o
ra
tive Beam
form
ing fo
r Multi-obje
c
tive Ra
diation… (N.
N
.N.A. Malik)
7325
Equation (4)
and (5
).
Step 7.
In order to form the
most pro
m
isi
ng perfo
rma
n
c
e of LAA:
Step 7a
. Rotate virtual line
in count
er clo
c
kwi
s
e
dire
ction with
different an
gle
as
s
h
ow
n
in
F
i
gu
r
e
3
.
Step 7b.
Return to
Step 1
for differe
nt angle rotatio
n
of virtual line
Step 7c
.
Co
mpare the
radiation
bea
mpattern
pe
rforman
c
e
re
sults from
d
i
fferent
rotations
.
Step 7d.
Select the be
st solution.
Step 8.
End
The reward
s
then a
s
soci
ated with thi
s
p
r
opo
se
d met
hodol
ogy are
the de
sign
o
b
jective
on the d
e
si
red radiation
beam
pattern with
mi
ni
mum SLL, d
e
sired m
a
in
beam
angle
and
expecte
d si
ze
of beamwidt
h in comp
ari
s
on with previous LFA [8].
Table 1.
List
of Paramete
rs used in WS
N Sche
me Im
plementatio
n of ILSA
Number of
nodes
Z
900
Region of interes
t
Λ
m
2
900
m
2
Number of
CB n
odes
N
(8,12,16
)
Communication r
adius
C
m
(6m, 8m, 14m
)
Angle of interest
(
0,
,
0
) (
0,
,
0
)
Table 2. List
of Paramete
rs and Valu
es
use
d
in HLPS
O Implement
ation
Number of
particles
S
30
Dimension of par
ticles
N
(8,12,16
)
Iterations
It
500
Range of pa
rticles
D
0 to 2
O
Upper bo
undar
y
for
d
n
U
N
2.2
O
Lo
w
e
r bo
undar
y
for
d
n
L
N
0.35
O
Upper bo
undar
y
for
d
1
U
1
2.2
O
Lo
w
e
r bo
undar
y
for d
1
L
1
0.3
O
Maximum uppe
r limit
U
max
0.1
O
Max
i
mum lo
w
e
r l
i
mit
L
min
2.5
O
Velocity
V
0 to 0.2
Learning factors
c
1
=c
2
2.0
Max
i
mum w
e
ight
ma
x
0.9
Minimum w
e
ight
min
0.4
4. Simulatio
n
and Anal
y
s
is
The key highl
ight in this n
e
w
algo
rithm i
s
t
he capa
bili
ty of selectin
g only the ap
prop
riate
colla
borative
active n
ode
s to pe
rform
as
an
array
antenn
a
whi
c
h m
eets th
e de
sired
m
u
lti-
obje
c
tives. T
herefo
r
e, in
o
r
de
r to de
mo
nstrat
e
the a
d
vantage
s of
the
propo
sed
ILSA for the
CB
beamp
a
ttern
perfo
rman
ce,
seve
ral
sim
u
lation
s in
m
u
ltiple
scena
rios, i.e. 8
-
, 1
2
-
and
16
-no
de
ILSA have b
een
con
d
u
c
te
d to p
r
ove th
e pe
rform
a
n
c
e of the
pro
p
o
se
d ILSA m
e
thod
com
p
a
r
ed
to previou
s
the LFA [8].
Ca
se 1
:
Case 1 investigates the
capability of ILSA to manage t
w
o desi
red objectives
simultan
eou
sl
y, i.e. SLL minimizatio
n
a
nd
ad
aptive
angle,
whi
c
h
are
ap
plied
on 8
-
no
de IL
SA
and
12
-nod
e
ILSA. An 8
-
node
ILSA is sim
u
lated
fo
r d
e
si
red
mai
n
be
am
angl
e of
-30
as
in
Figure
4. Re
sults are
co
mpared
to th
e co
rrespon
d
i
ng results o
b
tained
usi
n
g
LFA. It is sh
own
that the SLL
sup
p
re
ssion
i
s
imp
r
ove
d
.
All the
mino
r lobe
s h
a
ve
been
minimi
zed
with the fi
rst
SLL to be le
ss th
an -16
dB i.e. -16.5
9
dB and
th
e maximum
SLL is -15.3
0
dB. The si
ze of
FNBW of ILSA maintains
simila
rly as the FNB
W
of LFA, i.e. 43
. It can be inferred that the
result sho
w
s a very excelle
nt outcome.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 10, Octobe
r 2014: 731
8
– 7329
7326
Figure 4. Rad
i
ation Beamp
a
ttern of 8-n
o
de ILSA and 8-no
de LFA
The
next arra
y con
s
id
ere
d
is a
12
-n
ode
ILSA as
sh
o
w
n i
n
Fi
gure
5. A first
ob
servation
from these pl
ots is that a g
ood
pe
rforma
nce of radiati
on pattern is
obtaine
d
from
ILSA even with
a 30
adaptiv
e an
gle. F
o
r
a F
N
BW of 3
0
, the
SLL
a
c
hieve
d
i
s
-1
5.03 dB
a
s
compa
r
ed
to t
h
e
LFA, i.e. -12.68 dB. It can be inferred th
at
the results
sho
w
an a
c
ce
ptable out
co
me.
Figure 5.
Ra
diation Beam
pattern of 12
-node ILSA an
d 12-n
ode L
F
A
Ca
se 2:
In thi
s
case, 3 de
sired o
b
je
ctives of SLL
mini
mization, F
N
BW controll
a
b
le an
d
main b
eam d
e
sired
angle
are
co
nsi
dere
d
sim
u
ltaneo
usly a
s
referred to Fig
u
re
6. 16-nod
e IL
SA
is simul
a
ted
and an
alyze
d
in orde
r to prove the ca
p
a
bility of this prop
osed alg
o
rithm to han
d
le
multi-obj
ectiv
e
requi
reme
n
t
s. The
de
si
re
d mai
n
b
eam
angle
is poi
n
ted to
25
.
Th
e si
ze
of
F
N
B
W
by usin
g 16
-n
ode ILSA (i.e
. 52
) i
s
adj
u
s
ted to b
e
larger tha
n
that
obtaine
d u
s
in
g 16-nod
e
LF
A
(i.e. 27
). In addition, the maximum SLL for 16-n
od
e ILSA outperforme
d
LFA. The maximu
m
SLL of -1
6.38
dB ha
s be
en
achi
eved by
usin
g IL
SA a
s
comp
are
d
to high S
L
L b
y
using
LFA, i
.
e.
-13.16 dB. It can b
e
inferre
d
that the
result sho
w
s very excellent ou
tcome.
Figure 6. SLL Minimizatio
n
, FNBW Controll
able an
d M
a
in Beam An
gle Adaptabl
e
of 25
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Intelligent Lin
ear Collab
o
ra
tive Beam
form
ing fo
r Multi-obje
c
tive Ra
diation… (N.
N
.N.A. Malik)
7327
Ca
se 3:
In
this
ca
se, a
n
o
ther m
a
in
criteria i
s
add
ed which i
s
the null
pla
c
ement.
Therefore, four objectiv
es will be
counted as the
focal m
u
lt
i-objectives in t
e
rm
s of
(i)
SLL
minimization
(ii) F
N
BW co
ntrollabl
e (iii) main beam
desi
r
ed a
n
g
l
e and (iv) n
u
ll placement
s.
These four di
fferent cha
r
a
c
teri
stics are need
ed
to be
simultane
ou
sly achieve
d
. It can be se
en
from Fi
g. 7
th
at the
12-nod
e ILSA a
c
hi
e
v
es
ze
ro
sid
e
l
obe
s at
inten
ded
null
pla
c
ements of
-4
0°
and 4
0
°, while the si
delob
es of LFA
are un
cont
rolle
d and
high at
the intende
d
null pla
c
em
e
n
ts.
More
over, it can be
se
en th
at all other
mi
nor lo
be
s de
crea
sed
sig
n
ificantly with m
a
ximum SLL
of
-21.46 dB at 90° compa
r
e
d
to the higher sid
e
lo
b
e
s
of LFA with maximum SL
L of -12.88 d
B
at
17°. The
normalize
d
gain
demon
strated
maximum ga
in at
0
=0°. T
he FNB
W
of
12-n
ode ILS
A
is
then optimize
d
to be large
r
, i.e. 86
.
Figure 7. SLL Minimizatio
n
, FNBW Co
ntrollabl
e, Nul
l
Placeme
n
t and Main Bea
m
Angle
adapta
b
le of 0
Next, the main beam of this 12
-no
de ILSA is then steered to 25
as in Figu
re 8. As
comp
ared to
12-n
ode
LFA, ILSA achiev
es lo
we
r SLL
s
, wide
r F
N
B
W
up to 8
7
, nulling ability at -
100
and
10
0
an
d finally
adaptive
an
gle towards
25
. It is
obv
ious
ly noted
that this
newly-
prop
osed ILS
A
can
satisfy
these
four co
nflicts
by achi
eving the
opti
mum radiatio
n be
ampatte
rn
.
It can be inferred that the result is very e
x
cellent.
Figure 8.
SLL Minimizatio
n
, FNBW Controllable, Null Placem
ent and
Main Beam Angle
Adaptable of
25
Ca
se 4
: T
he
final array co
nsid
ere
d
is a
12-node
ILSA with differe
nt multi-obj
e
ctive of
spe
c
ified ran
ges of null
s
as de
picted i
n
Figure
9. The figure display
s
the re
sulting radiati
o
n
pattern
corre
s
po
ndin
g
to the othe
r two
desi
r
ed m
u
lti-obje
c
tive desi
gns of
control
l
ed FNB
W
an
d
minimum S
L
L
.
For
co
mpa
r
i
s
on
pu
rpo
s
e
s
, LFA is al
so
plotted in
this figure. It
can
be verifie
d
th
at
a good
perfo
rmance of ra
d
i
ation pattern
is attai
ned from 12-nod
e
ILSA as comp
ared to the
1
2
-
node LFA. In particula
r, ILSA has ze
ro
s (deep n
u
lls)
at the angle
s spe
c
ified to b
e
locate
d in the
rang
e
nu
[60
120
]. Note that all curv
es are enforced to
be low
values
with maximum SLL of -
15.29 dB. Th
e size of FNB
W
noticeably
increa
se
s up
to 50
.
Evaluation Warning : The document was created with Spire.PDF for Python.